Skip to content
aspens

Roadmap

Shipped

Section titled “Shipped”
  • Import graph (v0.2.0) — JS/TS + Python import parsing, file priority ranking, BFS domain clustering, git churn hotspots, tsconfig path alias resolution
  • Agentic discovery (v0.2.0) — Two parallel LLM agents (domains + architecture), findings fed into skill generation, parallel batches of 3
  • Vendored code exclusion (v0.2.1) — Skip vendor dirs, generated files, lock files
  • Auto skill activation (v0.2.2) — Hook-driven activation on every prompt, skill-rules.json with keyword/intent matching, session-sticky skills
  • Graph persistence (v0.3.0) — doc graph command, .claude/graph.json, code-map, graph-index, graph context hook
  • Full-repo refresh + custom skills (v0.4.0) — doc sync --refresh, add skill, interactive file picker, hardened git hook
  • Token optimizer (v0.5.0) — Prompt trimming across templates, plan + execute agent pair
  • Multi-target support (v0.6.0) — Codex as first-class target, target/backend separation, config in .aspens.json
  • Context health (v0.7.0) — doc impact command: freshness, domain coverage, drift, health score, LLM interpretation, interactive apply
  • Save-tokens (v0.7.0) — Token-saving hooks + handoff commands + session directory
  • Multi-language domains (v0.7.1–v0.7.2) — Domain discovery for C#, Java, Swift, PHP, Elixir, Kotlin, and F#; nested-project source-root detection; build-output skipping
  • triggers: skill frontmatter (v0.8.0) — Activation rules (files, keywords, alwaysActivate) in frontmatter; churn-stable code-map.md; dedicated Python/TS import parsers with alias resolution
  • Framework probing — Next.js (v0.8.0) — App Router, Pages Router, and special files (middleware, instrumentation) detected as implicit graph roots
  • OpenCode support (v0.9.0) — opencode as a first-class backend and target (AGENTS.md + .claude/skills); .php counted as source
  • Bundled agents — 11 agent templates, aspens add, customize agents, --recommended install

Next up

Section titled “Next up”
  • More framework probing — Django, Rails, Spring Boot, Laravel architecture detection
  • Multi-language import graphs — hub files and clustering beyond JS/TS/Python (C#, Java, Go, Rust, etc.)
  • Infrastructure detection — CI/CD, migration tools, API specs, IaC
  • Symbol extraction — web-tree-sitter for function/class/type signatures, Aider-style repo map
  • Better clustering — Louvain community detection, temporal coupling from git, PageRank
  • Workspace detection — pnpm/npm/yarn/nx/turbo monorepo topology

Future

Section titled “Future”
  • Direct API mode — --backend api for CI/CD without a coding-agent CLI installed
  • Semantic search (maybe) — aspens doc search with local vector store